Development and internal validation of diagnostic prediction models using machine-learning algorithms in dogs with hypothyroidism.

Introduction: Hypothyroidism can be easily misdiagnosed in dogs, and prediction models can support clinical decision-making, avoiding unnecessary testing and treatment. The aim of this study is to develop and internally validate diagnostic prediction models for hypothyroidism in dogs by applying machine-learning algorithms. Methods: A single-institutional cross-sectional study was designed searching the electronic database of a Veterinary Teaching Hospital for dogs tested for hypothyroidism. Hypothyroidism was diagnosed based on suggestive clinical signs and thyroid function tests. Dogs were excluded if medical records were incomplete or a definitive diagnosis was lacking. Predictors identified after data processing were dermatological signs, alopecia, lethargy, hematocrit, serum concentrations of cholesterol, creatinine, total thyroxine (tT4), and thyrotropin (cTSH). Four models were created by combining clinical signs and clinicopathological variables expressed as quantitative (models 1 and 2) and qualitative variables (models 3 and 4). Models 2 and 4 included tT4 and cTSH, models 1 and 3 did not. Six different algorithms were applied to each model. Internal validation was performed using a 10-fold cross-validation. Apparent performance was evaluated by calculating the area under the receiver operating characteristic curve (AUROC). Results: Eighty-two hypothyroid and 233 euthyroid client-owned dogs were included. The best performing algorithms were naive Bayes in model 1 (AUROC = 0.85; 95% confidence interval [CI] = 0.83-0.86) and in model 2 (AUROC = 0.98; 95% CI = 0.97-0.99), logistic regression in model 3 (AUROC = 0.88; 95% CI = 0.86-0.89), and random forest in model 4 (AUROC = 0.99; 95% CI = 0.98-0.99). Positive predictive value was 0.76, 0.84, 0.93, and 0.97 in model 1, 2, 3, and 4, respectively. Negative predictive value was 0.89, 0.89, 0.99, and 0.99 in model 1, 2, 3, and 4, respectively. Discussion: Machine learning-based prediction models were accurate in predicting and quantifying the likelihood of hypothyroidism in dogs based on internal validation performed in a single-institution, but external validation is required to support the clinical applicability of these models.

Web Health Monitoring Survey: A New Approach to Enhance the Effectiveness of Telemedicine Systems.

BACKGROUND: Aging of the European population and interest in a healthy population in western countries have contributed to an increase in the number of health surveys, where the role of survey design, data collection, and data analysis methodology is clear and recognized by the whole scientific community. Survey methodology has had to couple with the challenges deriving from data collection through information and communications technology (ICT). Telemedicine systems have not used patients as a source of information, often limiting them to collecting only biometric data. A more effective telemonitoring system would be able to collect objective and subjective data (biometric parameters and symptoms reported by the patients themselves), and to control the quality of subjective data collected: this goal be achieved only by using and merging competencies from both survey methodology and health research. OBJECTIVE: The objective of our study was to propose new metrics to control the quality of data, along with the well-known indicators of survey methodology. Web questionnaires administered daily to a group of patients for an extended length of time are a Web health monitoring survey (WHMS) in a telemedicine system. METHODS: We calculated indicators based on paradata collected during a WHMS study involving 12 patients, who signed in to the website daily for 2 months. RESULTS: The patients' involvement was very high: the patients' response rate ranged between 1.00 and 0.82, with an outlier of 0.65. Item nonresponse rate was very low, ranging between 0.0% and 7.4%. We propose adherence to the chosen time to connect to the website as a measure of involvement and cooperation by the patients: the difference from the median time ranged between 11 and 24 minutes, demonstrating very good cooperation and involvement from all patients. To measure habituation to the questionnaire, we also compared nonresponse rates to the items between the first and the second month of the study, and found no significant difference. We computed the time to complete the questionnaire both as a measure of possible burden for patient, and to detect the risk of automatic responses. Neither of these hypothesis was confirmed, and differences in time to completion seemed to depend on health conditions. Focus groups with patients confirmed their appreciation for this "new" active role in a telemonitoring system. CONCLUSIONS: The main and innovative aspect of our proposal is the use of a Web questionnaire to virtually recreate a checkup visit, integrating subjective (patient's information) with objective data (biometric information). Our results, although preliminary and if need of further study, appear promising in proposing more effective telemedicine systems. Survey methodology could have an effective role in this growing field of research and applications.

Is the digital divide an obstacle to e-health? An analysis of the situation in Europe and in Italy.

INTRODUCTION: The digital divide affecting elderly patients may compromise the diffusion of telemedicine systems for this age segment. It might be that the difficulties in the passage from trials to the effective distribution of telemedicine systems are also due to the awareness of a personal digital divide in the target population. MATERIALS AND METHODS: The analysis aims to estimate the number of people over the age of 50 years with potential cardiovascular problems able to access the Web. It made use of data from several sources (the Survey of Health, Ageing and Retirement in Europe and the Istituto Nazionale di Statistica Multiscopo Survey). Furthermore, with regard to Italy, the estimates obtained from official data were compared with those obtained in a survey investigating heart failure patients in Tuscany. RESULTS: In 2011, the percentage of people suffering from cardiovascular diseases and with Web access was 24% in Europe, with significant differences by country (ranging from 53% in Switzerland to below 20% in Italy, Spain, and Portugal). In Italy, however, the proportion of people with Web access increased from 2007 to 2011, and the survey in Tuscany showed that elderly people with limited information and communications technology skills overcame challenges and learned how to connect to the Web because they started to appreciate new technologies. CONCLUSIONS: The opportunity to use the Internet to monitor patients with chronic disease can serve as a challenge to reduce the digital divide gap and, furthermore, to increase their social and technological inclusion.

The informative contribution of the "virtual medical visit" in a new heart failure telemedicine integrated system.

INTRODUCTION: Telemedicine systems consist of collection, transmission, and analysis of biometric data essentially based on instrumental measures. Our goal was to evaluate if information collected from patients has an incremental informative value in automatically rating the patient's health status. MATERIALS AND METHODS: We present preliminary results of a new telemedicine system (ASCOLTA) obtained by observation of 12 heart failure patients (New York Heart Association Class IIb-III). Instrumental data (electrocardiogram, oxygen saturation level, and respiration rate) were wirelessly collected daily together with clinical data (weight, heart rate, and blood pressure values) and patients' information obtained through a Web-based questionnaire, simulating a virtual medical visit. Health status was independently judged by two blinded cardiologists and by the patient's cardiologist on the basis of 348 daily clinical reports. Random forest classification analysis was applied to 240 complete clinical report variables in order to estimate the judged health status. RESULTS: The use of "patient's information" led to a better predictive ability in comparison with using only physiological parameters assessed by instruments. The complete set of variables (Patient+Instrumental) achieved 84% concordance, compared with 72% for the instrumental-only variables and 69% for the patient-only variables. The receiver operator characteristics curves graphically confirmed the described results. CONCLUSIONS: Patients have an active role in home monitoring, and their information appears relevant for a new telemedicine approach integrating subjective and objective vital signs. Combining patient information with instrumental parameters, it is possible to achieve a more correct automatic classification of health status of heart failure patients.